import torch
import io
from pathlib import Path
from fastapi import FastAPI, File, UploadFile, Form
from fastapi.responses import HTMLResponse
from PIL import Image
from torchvision import transforms
import uvicorn
from train import BoneAgeModel
app = FastAPI(title="Bone Age Assessment")
device = torch.device("cpu") # CPU only for Render free tier
model = None
use_fp16 = False
# Try fp16 model first (smaller memory), fallback to full model
for model_file in ["best_model_fp16.pth", "best_model.pth"]:
model_path = Path(model_file)
if model_path.exists():
m = BoneAgeModel()
if model_file == "best_model_fp16.pth":
m = m.half()
use_fp16 = True
m.load_state_dict(torch.load(model_path, map_location="cpu"))
m.eval()
model = m
print(f"Model loaded: {model_file}")
break
if model is None:
print("No trained model found")
transform = transforms.Compose([
transforms.Resize((256, 256)),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
def predict_bone_age(image: Image.Image, is_male: bool) -> float:
img_tensor = transform(image.convert("RGB")).unsqueeze(0).to(device)
gender_tensor = torch.tensor([1.0 if is_male else 0.0]).to(device)
if use_fp16:
img_tensor = img_tensor.half()
gender_tensor = gender_tensor.half()
with torch.no_grad():
prediction = model(img_tensor, gender_tensor)
return prediction.item()
@app.get("/", response_class=HTMLResponse)
async def home():
return """
Bone Age Assessment
Bone Age Assessment
Upload a hand X-ray to estimate skeletal maturity
Developed by Dr. Prashanth Mamidipalli
Estimated Bone Age
⚠️ For research and educational use only. Not intended for clinical diagnosis or patient management.
"""
@app.post("/assess")
async def assess(
file: UploadFile = File(...),
sex: str = Form(...)
):
if model is None:
return {"error": "Model not loaded. Please train the model first."}
contents = await file.read()
# Basic validation
if len(contents) < 10 * 1024: # less than 10KB
return {"error": "File too small. Please upload a valid hand X-ray image."}
image = Image.open(io.BytesIO(contents))
# Check if image is grayscale-ish (X-rays have low color saturation)
img_rgb = image.convert("RGB")
import numpy as np
arr = np.array(img_rgb).astype(float)
r, g, b = arr[:,:,0], arr[:,:,1], arr[:,:,2]
color_diff = np.mean(np.abs(r - g) + np.abs(g - b) + np.abs(r - b))
if color_diff > 30:
return {"error": "Image appears to be a color photo, not an X-ray. Please upload a hand X-ray image."}
is_male = sex.lower() == "male"
predicted_months = predict_bone_age(image, is_male)
years = int(predicted_months // 12)
months = int(predicted_months % 12)
years_decimal = round(predicted_months / 12, 2)
return {
"bone_age": f"{years} yrs {months} mo",
"predicted_months": round(predicted_months, 1),
"predicted_years": years_decimal,
"predicted_formatted": f"{years} years {months} months",
"total_months": round(predicted_months, 1)
}
if __name__ == "__main__":
uvicorn.run(app, host="0.0.0.0", port=8000)